fix: route FalkorDB queries to correct graph for single group_id

[spencer2211]

Fixes #1325

Summary

The handle_multiple_group_ids decorator only switched the FalkorDB driver to the correct graph database when len(group_ids) > 1. For a single group_id (the most common case), queries fell through to the default execution path, hitting default_db which typically has no data. This caused silent empty results for all FalkorDB deployments using per-query group_ids.

Additionally, all get_by_group_ids classmethods on node and edge types bypassed the decorator entirely, querying the driver directly without routing to the correct FalkorDB graph.

Changes

decorators.py

• Change len(group_ids) > 1 to len(group_ids) >= 1 so the decorator activates for any non-empty group_ids list

nodes.py (4 methods)

• Add FalkorDB graph routing to EpisodicNode.get_by_group_ids, EntityNode.get_by_group_ids, CommunityNode.get_by_group_ids, SagaNode.get_by_group_ids

edges.py (5 methods)

• Add FalkorDB graph routing to EpisodicEdge.get_by_group_ids, EntityEdge.get_by_group_ids, CommunityEdge.get_by_group_ids, HasEpisodeEdge.get_by_group_ids, NextEpisodeEdge.get_by_group_ids

Approach

For each get_by_group_ids method:

• Single group_id: Clone the driver to target the group's graph database (driver.clone(database=group_ids[0])), then fall through to the existing query logic
• Multiple group_ids: Query each graph separately via recursive call with a single group_id, merge results, re-sort by UUID descending, and apply limit

This approach avoids code duplication — the single-group case reuses the existing query path, and the multi-group case decomposes cleanly into single-group calls.

Context

FalkorDB uses separate graph databases per group_id, unlike Neo4j which stores all data in a single database and filters by group_id as a node/edge property. The decorator and classmethods need to explicitly switch the driver to the correct graph database before executing queries.

Discovered and verified against a production FalkorDB deployment running graphiti-core with the graphiti MCP server.

[danielchalef]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[spencer2211]

I have read the CLA Document and I hereby sign the CLA